Proceedings of the 8th Conference on Architectural Support for Programming Languages and Operating Systems, 1998. A Cost-Effective, High-Bandwidth Storage Architecture Garth A. Gibson*, David F. Nagle†, Khalil Amiri†, Jeff Butler†, Fay W. Chang*, Howard Gobioff*, Charles Hardin†, Erik Riedel†, David Rochberg*, Jim Zelenka* School of Computer Science* Department of Electrical and Computer Engineering† Carnegie Mellon University, Pittsburgh, PA 15213
[email protected] ABSTRACT width, that is, linearly increasing application bandwidth with increasing numbers of storage devices and client processors, This paper describes the Network-Attached Secure Disk the data must be striped over many disks and network links (NASD) storage architecture, prototype implementations of [Patterson88]. With 1998 technology, most office, engineer- NASD drives, array management for our architecture, and ing, and data processing shops have sufficient numbers of three filesystems built on our prototype. NASD provides scal- disks and scalable switched networking, but they access stor- able storage bandwidth without the cost of servers used age through storage controller and distributed fileserver primarily for transferring data from peripheral networks bottlenecks. These bottlenecks arise because a single (e.g. SCSI) to client networks (e.g. ethernet). Increasing “server” computer receives data from the storage (periph- eral) network and forwards it to the client (local area) dataset sizes, new attachment technologies, the convergence network while adding